Conference Agenda

This study addresses a critical methodological gap in social network analysis by developing a novel analytical framework that adapts the "prescribing cascade" concept from medical science to analyze evolutionary complexity in R&D networks in science and technology policy implementation. While existing SNA approaches capture network structure at discrete timepoints, they lack tools for analyzing cascading effects of policy interventions across multiple stages of network evolution.

The proposed mixed methods framework integrates four matrix-based analytical techniques: (1) Basic adjacency matrices capturing initial R&D network relationships, (2) Weighted adjacency matrices quantifying relationship directionality across intervention stages, (3) Policy intervention matrices tracking new institutional relationships introduced by each policy intervention, and (4) Network evolution matrices measuring changes in key network properties including density and average path length across cascade stages. We operationalize this framework by developing weighted matrices that capture the nature of relationships between actors (universities, industries, government agencies, coordination bodies, and evaluation institutions), tracking from Stage 0 (initial network) through Stage 1 (first cascade) to Stage 2 (second cascade).

Our methodological contribution demonstrates how tracking matrix transformations and resulting network property changes can reveal cascade patterns in policy interventions. This matrix-based methodology offers researchers a systematic way to analyze how initial policy interventions designed to enhance R&D coordination can paradoxically increase network complexity. Beyond R&D networks, the framework can be applied to various complex systems where policy interventions trigger institutional adaptations, advancing both theoretical understanding of policy-induced network evolution and practical methodological tools for analyzing cascade effects in social networks.

The collection and analysis of empirical temporal contact networks have experienced remarkable growth over the past two decades. Sociopatterns, a research collaboration, has gathered high-temporal resolution data on physical proximity and face-to-face interactions across a wide range of social contexts (J. Stehlé et al., 2011). These datasets have paved the way for a new wave of quantitative studies on individual and social behaviors. This advancement is particularly significant when combined with sociological and psychological metadata collected through participant surveys, as demonstrated in studies conducted during conferences (M. Génois et al., 2019).

Disentangling phenomena that require social explanations from those that do not is a complex task, necessitating the development of new analytical methods. Further research has shown that certain behavioral characteristics appear to be universal and can be explained by simple mechanisms (R. Masoumi et al., 2024). In this context, we identify among all determinants measured onto the empirical face-to-face networks, the ones that can be correlated with social and psychological behavior, from others with universal properties. Moreover, by leveraging the diversity of social contexts studied using consistent data collection methods, alongside statistical tests and renormalization processes to assess the relevance of behavioral observables, we compare these determinants across different contexts through longitudinal studies. For instance, this method unveils that loyalty, the repetition of interaction, is correlated to individual properties and social context.

This method, by integrating network features with metadata analysis, provides researchers with a more comprehensive framework for analyzing and explaining social behaviors in face-to-face interaction networks.

Aims: We sought to understand how structural configurations of personal social networks collectively explain perceived social support among older adults with advanced cancer. We used Qualitative Comparative Analysis (QCA), a systematic approach for comparing multiple cases to discern how different combinations of factors jointly produce an outcome. Rather than isolating single variables, QCA treats each condition as a set and uses set-theoretic principles to reveal which combinations of set memberships reliably account for the outcome. Designed for small to medium samples, QCA uncovers real-world complexities overlooked by conventional statistical approaches.

Methods: Fifteen older adults with advanced cancer participated in the study. Each participant was guided through completing a personal network chart—a visual map showing the key individuals involved in their health and well-being (1). On this chart, participants specified how different people (e.g., spouse, family, friends, neighbors) connect both to them and to one another. They also completed the Berkman-Syme Social Network Index, a tool that measures perceived social support.

We conducted a fused Mixed-Methods analysis (2,3), integrating insights from personal network visualizations (qualitative data) with quantitative network analysis. This approach enabled us to identify various structural patterns within personal networks. We then applied QCA to determine which sets of configurations predict higher social support. For QCA, we established four theoretically grounded conditions: (1) existence of a network core (one or more crucial individuals—such as a spouse—connected to many others), (2) existence of a dominant cohesive circle (a tightly interconnected cluster, for instance, of family members), (3) the presence of segregated clusters or isolated individuals (e.g., friends, colleagues, extended family), and (4) the size of the inner circle in the network chart. QCA helped us pinpoint consistent combinations of these conditions that lead to higher perceived social support.

Using the intermediate solution—a QCA approach that employs theoretically guided assumptions while avoiding extreme simplifications—we found that patients tend to perceive higher social support when any of the following configurations is present: (1) the network is less segregated, featuring fewer isolated clusters and more interconnected ties; (2) a spouse is part of the network alongside a larger inner circle; or (3) the network has a larger inner circle even without cohesive clusters. These configurations demonstrated high consistency of 0.85 (measuring how reliably a set of conditions is associated with the outcome, Pi consistency: 0.81) and coverage of 0.81 (the proportion of outcome cases that a particular set of conditions explains).

In addition, necessity analysis showed that most instances of higher support had either a less segregated network or a larger inner circle (relevance of necessity: 0.553, necessity coverage: 0.783).

Conclusion: This study demonstrates that less segregated networks or the existence of a robust inner circle predict a higher sense of social support among older adults with advanced cancer. Notably, even networks without overall cohesion patterns still achieve high levels of support with a robust inner circle. These findings can inform the adaptation of network-building interventions to reduce isolation and strengthen social support.